what makes a data scientist great? · pdf filewhat makes a data scientist great? ... a big...

SPECIAL COLLECTION • “[TITLE]”• MIT SLOAN MANAGEMENT REVIEW i

CONTENTSSPECIALCOLLECTIONFALL 2015

What Makes a Data Scientist Great?

1 General Mills Builds Up Big Data to Answer Big Questions Wayde Fleener (General Mills), interviewed by Michael Fitzgerald

7 Business Quandary? Use a Competition to Crowdsource Best Answers Anthony Goldbloom (Kaggle Inc.), interviewed by Renee Boucher Ferguson

13 You Need a Data Dictator? Jeanne Ross (MIT Sloan), interviewed by David Kiron and Renee Boucher Ferguson

General Mills Builds Up Big Data to Answer BigQuestionsWAYDE FLEENER (GENERAL MILLS), INTERVIEWED BY MICHAEL FITZGERALD

Enhancing intuition with analytics at General Mills.

At General Mills, a big data team within the Consumer Insights group

uses data visualization to help share their insights. "Now the divisions

can actually tangibly understand what they’re doing from a data

perspective," says Wayde Fleener, a data scientist and Senior Manager in

the Global Consumer Insights group.

Wayde Fleener likes solving real-world problems, which

led him into what we now call Big Data analytics. He

started out on a different path, studying chemical

engineering at the University of Minnesota. But he

realized he didn’t want to go on for a PhD because it

would mean a career in research, and he wanted to do

applied work. After a stint in investment banking in Asia,

Fleener went to a loyalty marketing agency.

Fleener, 36, still remembers the day when a colleague

came to him and said a name, “big data,” had been coined

for what they were doing. “We called it decision science,”

he says. After working on programs for the consumer

packaged goods industry, he found himself intrigued by

the challenges they face, as companies indirectly

connected to their consumers. He came to General Mills

in mid-2013, becoming a senior manager/data scientist in

its 200+-person Consumer Insights group, a unit of

marketing comprised largely of researchers. He spoke

with MIT Sloan Management Review contributing editor

Michael Fitzgerald.

What did General Mills want in a datascientist for its marketing group?

The leader for consumer insights, Jeanine Bassett [vice

president of Global Consumer Insights at General Mills]

was concerned about how much of our decision making

was based on outside research. She wanted more decision

making based on our intuition and utilizing the data we

had in-house. She wanted us to become less reliant on

research. During the whole interview process, they said,

MITMIT SLSLOOAN MANAAN MANAGEMENGEMENT REVIEWT REVIEW

DDAATTA & ANALA & ANALYTICSYTICS

Copyright © Massachusetts Institute of Technology, 2015. All rights reserved. Reprint #56424 http://mitsmr.com/1G7LlKXSPECIAL COLLECTION • “[TITLE]”• MIT SLOAN MANAGEMENT REVIEW 1

“We need someone to really help us to answer these

questions, about what we think we know.” So I came in to

get after the data that we had internally and how we could

benefit from it. And they gave me a lot of leeway.

Did they also give you a sense of urgency, ofneeding to get things done quickly?

Yes. They wanted everything as quickly as possible but

not at the sacrifice of accuracy.

How did you get started?

Coming in as an outsider, my first couple of months were

focused on figuring out what we really do at General

Mills. The very first thing I had to do was sit down and do

a catalog of all the different data we commonly use and

what we use it for. I started asking around, “Does

anybody have a data diagram that tells me the data you

have and where it’s being used?” No one had it. So I

started doing diagrams to say where the data was

connected, where are they not, and asking, where should

they be? Where are the various data being used in

analyses? Where should they be?

How did you find the data?

I had to do a bunch of interviews. I found the group that

does our marketing mix, and sat down with them for

quite a while. I went to our relationship marketing group,

and asked them what data they were using. I’d sit down

with all these specialized functions inside Consumer

Insights and ask a lot of questions.

Then I’d get really tactical, get access to the data sources

themselves, and literally go in and look at tables — there

are no data dictionaries. I had to figure out what [did] the

columns inside the tables mean. A lot of times, I had to

work my way through the IT organization, because the

only person who had the answer was the person who

developed that database. Sometimes there were names for

columns and I couldn’t interpret what they meant.

After that, I had to figure out where the data was coming

from upstream. And I realized there was data coming to

us that we weren’t capturing that we should be capturing.

We would just delete it.

What kind of data was not being captured?

For example, we may bring it into our system at an

aggregated market level, but the data is actually coming to

us at a disaggregated level [by individual locales].

What was the most difficult thing at thebeginning?

General Mills is a huge ship, and it’s very difficult to

move. People felt threatened because I was coming into

their role and questioning them. So there was a lot of

resistance, and “Are you trying to take over what I’m

doing?” I had to keep saying, “No, I’m just trying to help

you do things better.” I sit in the business, but my role

goes into IT a lot. So the very first couple of months, there

was tension between me and IT, because they were afraid

they were going to be told they had been doing things

wrong. I resolved it through weekly meetings, showing

them what we’re doing; not calling people.

Did you do this all yourself, or did uppermanagement need to get involved?

There were a couple of instances where we hit roadblocks

and there had to be a senior management talks. It took six

months before we really turned a corner. One of the

director-level leaders in IT started one of the meetings

saying, “Hey, listen,” — and this was directed to some of

their own folks — “Consumer Insights has some really

interesting things that I believe we’re going to benefit

from as a company.” And then the leader left the room.




When that meeting happened, I knew the tide had

changed.

Were you working with anyone else?

I started off as a one-man team. As people started

approaching me, my VP saw the traction and invested a

little more, and we were able to show value, and it just

kept on growing. Once we had a sizeable amount of our

data connected, there was a sense of urgency to show the

value of that. That is why we jumped into visuals so the

organization could tangibly touch something that we did.

Data visualization really accelerated things. We were

sorting out a lot of data, and behind the scenes the data

was coming together, but a lot of the Consumer Insights

folks couldn’t tangibly touch it yet. We sat down with

some of our division folks and identified the things they

wished they could do but couldn’t because they couldn’t

access the data. I found a data-visualization company that

could take the data sets we were bringing together and

produce some valuable visuals. Now the divisions can

actually tangibly understand what they’re doing from a

data perspective.

That just grew to more requests and more people who

wanted visualizations. We’re now the Big Data team

within Consumer Insights. Our mix after a little more

than a year is myself, four data stewards, their manager,

four data visualizers, and two data scientists. We need two

more data scientists, something I’m actively working on

now.

Was data visualization a new role for GeneralMills?

When I got here, General Mills already had in place a

data visualization tool. But it was not used at all in

Consumer Insights. My researchers, who are not

programmers, live in Excel, but my datasets are hundreds

of millions of rows. [Editor’s note: At the time, Microsoft

did not have a version of Excel that could handle such

large numbers of rows. General Mills has since become a

beta tester of a version that does.] With our visualization

tool, we can do a lot of stuff out of the box without having

to program, but if you’re trying to do specialized stuff

then you need to bring in programming languages. I

started looking for some developers that knew our tool

and could do the more advanced stuff. We outsourced

that data visualization to another company.

Data visualization technology is now a commodity. But,

the idea of what you need to visualize is not. That’s what

we data scientists do. We come up with “it’d be really cool

if we could do this” ideas. When we’re ready to productize

it, the programming is the commodity piece.

What are the pros and cons of looking outsidefor talent versus staying inside?

Outside, you get the advantage of individuals that have

developed across industries and other companies. You

may create a blueprint and hand it to them, and they

come back and say we developed something similar. Here

are some ideas that might enhance your visualization.

Internally, it always comes down to cost: can you bring it

in-house and do it cheaper? Some of our outsourcing is

through India, and it can be almost like having 24-hour

development.

You’ve talked about how to make analyticspart of everyday operations, not havinganalytics be a series of one-off projects. Howhard is that process to develop?

It’s very difficult. People are living in an Excel world. They

think, “I gotta grab this data and this data and every time




I want to refresh.” Versus, we’re going all the way back to

the data source in everything that we’re doing, so we don’t

have to touch that whole data update.

If analytics traditionally works in a rinse-and-repeat mode, where you ask a question,gather data, run the data, ask anotherquestion, gather new data, run it, and so on,how do you break away from that?

Every division thinks it has this unique way it should

view the data. Yes, they all face their own unique

situation, but in reality, they really look at the data the

exact same way. If I have to create a unique way for you to

view data across seven divisions, I have to maintain seven

unique versions of the data. I can’t get scale that way.

Does that help you avoid what you’ve calledthe “Go Fish” problem of Big Data, where youask people if they have the data, and they sayno?

When I first got here, no one had a view of all the data we

had across everything. Everybody had a specific role and

they cared about the data they used in that role. If

someone had a new use-case they developed, you would

have to go around and say, do we even have that data?

How do I get the data, how do I manipulate the data, how

do I combine it with my other data sets? When I mash up

my data sets, do I know I have yellow-box Cheerios

named consistently across the two data sets?

How did you fix that problem?

That’s where the data stewards, a new role within

Consumer Insights, come into play. We’ve been

systematically going through every single core data asset.

We’ve got hundreds of products. One data source may

spell “Cheerios” all the way out, and another may spell it

“chrs.” We’ll prepare a summarized data set and give that

to our data stewards. We have about six data stewards that

map connections between data sets. For every data

refresh cycle, we created what we call an exception report,

where we identify what’s new. The data stewards go

through and map what’s new.

Were the data stewards there before you gotto General Mills, perhaps doing somethingelse?

It’s grown three-fold. We had an individual solely focused

on Nielsen scanner data, a very important data set in any

consumer packaged goods company, making sure that

whatever we get from Nielsen will work the way

marketing needs it. The job had a different title. But now

they’re data stewards, and we’re taking those roles and

instead of making them solely an expert in one data

source, we’re making them an expert across data sources.

They reported to different lines when I got here, but as of

January they report to me. Data stewards don’t require all

the technical and programming skills of a data scientist.

For some of their tasks, a two-year technical degree can

be enough.

When and how did you add your next datascientist?

After six months, the business was asking for more

analytics, especially with a predictive component. The

amount of stuff that they were asking for was exceeding

my capacity. We’ve got a good recruiting office, so I built

out the profile I was looking for, and they started looking

at various academic programs locally. I think they found

him on LinkedIn, because his profile said he was studying

Big Data at The University of St. Thomas. I had told HR

that I was fine with an individual that was mid-school,

because that was probably what I could afford. He had a




day job already, and I had to do a big selling job to get

him. They’re tough to acquire. The big sell was, “We’re

starting something here at General Mills, you could be

part of it.” That’s appealing to certain individuals.

Your unit also has an analytics group withstatisticians. How are they different?

They don’t contain the data scientists. The data science

group is more technically savvy at getting and

manipulating the data, and can engineer the systems to

operationalize the analytics. The statisticians/analytics

group focuses on answering particular questions, but for

example, they do not have the skillset to take a model and

deploy it. Both groups report to Vidyotham Reddi, who is

director of Global Enablers. He reports to Jeanine Bassett.

How you can tell the company is successfullyoperationalizing analytics?

For us, it’s how many people are using our stuff and

asking for more stuff. Ultimately, a company is successful

if it can say decisions made by analytics led to

incremental profit.

What lessons do you have for othercompanies?

Start with an understanding of the data you have, how it’s

organized. Have the end in mind — what you want to

accomplish — and then develop that plan through the

process. Build the right relationships. I identified who

would be my champions and built relationships with

them.

You’ve got to make sure you have the right IT folks along

for the ride. I’m technical, but I hit a point where I don’t

understand Linux servers, and so I need an expert.

Having a breadth of relationships across the company is

the only reason why we have been successful.




About the Author

Michael Fitzgerald is a contributingeditor at MIT Sloan ManagementReview.




Business Quandary? Use a Competition toCrowdsource Best AnswersANTHONY GOLDBLOOM (KAGGLE INC.), INTERVIEWED BY RENEE BOUCHER FERGUSON

Top data scientists often share three characteristics: they are creative, they are

curious and they are competitive. Anthony Goldbloom, CEO of Kaggle, a company

that hosts data prediction competitions, has figured out how to tap all three of these

characteristics to help companies crowdsource their analytics problems.

Anthony Goldbloom, CEO of Kaggle

During a 2008 internship at the Economist, economist-

turned-journalist Anthony Goldbloom made an

important discovery: many organizations that want to use

predictive analytics don’t have the skills to do it well, if at

all. In response, he came up with an interesting business

idea: create a company that allows anyone anywhere to

compete to solve other business’s analytics problems.

Today, hundreds of organizations, both public and

private, have submitted their datasets and business

problems to Goldbloom’s company, Kaggle Inc. Sixty five

thousand data scientists from around the globe have

signed up with Kaggle to compete to develop the best

algorithm for a given problem, from predicting click-

through rates on ads to predicting who will be admitted

to a hospital within the next year (a current competition

with a $3 million purse). Ongoing results are displayed in

real time on a virtual leaderboard.

The leading data scientists are a motley crew. They

include a French born actuary, a computational

neuroscientist from Harvard, an Oxford physicist and a

Russian cybernetics professor.

In a conversation with MIT Sloan Management Reviewdata & analytics contributing editor Renee Boucher

Ferguson, Kaggle founder and CEO Goldbloom talks

about the motivation behind Kaggle and the outcomes of

its data science competitions.

SSo wo whherere did te did thhe idee idea coa comme fe frroom tm to mo maakke de daatta sa scienciencece

inintto a so a sppoorrts cots commppetetiittioion?n?

I used to be an econometric modeler at the Australia

Treasury, and after that at the Reserve Bank of Australia.

Then in early 2008, I won an essay competition. The prize

was a three-month internship at the Economist magazine,

writing for the finance and economic section.

I pitched a piece on predictive analytics and it turned out

to be a fabulous way to do market research. Pretty senior

people were saying, “yes, predictive modeling, I know we

need to be doing more of this.” And my frustration was



Copyright © Massachusetts Institute of Technology, 2012. All rights reserved. Reprint #54222 http://mitsmr.com/1k6rQoYSPECIAL COLLECTION • “[TITLE]”• MIT SLOAN MANAGEMENT REVIEW 7

that I was talking through some really interesting

problems with reasonably senior people from companies,

and always in the back of my mind was, “wow, I do a bit

of programming. I can understand business problems and

also like playing with data and statistics. I could do a

really good job on these problems.” But I realized if in a

different context I applied for a job to tackle some of

these issues, I wouldn't get one. They'd look at me and

say, “you're an economist, what do you know about

predictive modeling?”

That was the emotional rationale behind wanting to build

a meritocratic labor market for data science and

predictive models. Because I figured there would be more

people like me who thought they could solve some of

these problems, but they weren't being given the chance.

There was also an intellectual, or commercial, reason why

I thought it was a good idea. The quality of the data

scientist is crucial in determining the quality of a data

science solution. Data science is extremely high leverage,

so a data set and an algorithm in the hands of a great data

scientist can generate a $100 million return on

investment, while the same data in the hands of

somebody inexperienced can be extremely costly.

HHoow do Kw do Kaagggglele’’s bs big dig daatta coa commppetetiittioionns ws woorrk?k?

We basically back-test people's algorithms on what we

call a holdout data set.

So let's say you're a bank and you want to predict who's

going to default on a loan and you've got 20 years' worth

of data on how loan applicants have behaved, who have

paid back their loans and who hasn't. What we do is we

take that historical data, split it in two, give data scientists

in our competition access to half of that data set with the

answers, like “John Smith, age 62, with an income of X,

defaulted.” For the other half of the customers, we give

information on their characteristics — “Jack Smith, age

61, with income X” — but we withhold the information

on whether or not he defaulted.

What participants do is train their algorithms on the first

set, and then apply those algorithms to the second set,

and we score them against the actual answers to see who's

getting the lowest error rate or the highest number of loan

applicants correct. That's what's very neat about data

science and predictive modeling algorithms. You can

actually objectively score who is good and who isn't.

WWhho ao arre te thhe de daatta sa sciencienttiissts tts thhaat pt paarrtticiicippaatte in ye in yoour pur puubbliclic

cocommppetetiittioionns as annd hd hoow cw caan yn yoou tu teelll wl whho doo does ges gooood wd woorrkk

aannd wd whho doo doesnesn’’t?t?

We have 65,000 data scientists signed up to Kaggle, and

just like with golf tournaments, we have them all ranked

from 1 to 65,000. And the top data scientists come from a

whole lot of backgrounds. Number one is a French-born

actuary named Xavier Conort who runs his own

consulting firm. Number two is an Oxford PhD in physics

called Jason Tigg. He runs his own hedge fund. As an

intellectual challenge, he competes in Kaggle

competitions at night. Number three is a professor of

cybernetics from Moscow State University called

Alexander D'yakonov. Number four, Sergey Yurgenson, is

a professor of computational neuroscience at Harvard

Medical School.

The background of these top data scientists is all over the

map, but they have a handful of characteristics in

common. They are all creative and curious. In order to

build a really good data science algorithm, it's actually

much less about the complex mathematics. You have to




be really good at math, but the complexity of the

mathematics doesn't very well correlate with how good a

data science algorithm is. It's more about the kind of

features you're able to find in the data.

ThThaatt’’s rs reeaalllly iny inttereresesttiningg. C. Caan yn yoou gu giivve ae an exan exammpple?le?

Sure. We ran a competition for a big chain of used car

dealers. They sell used cars all over the country. They'll go

to auctions of secondhand cars, purchase cars and sell

them throughout the dealer network. Because they offer a

warranty on any car they sell — and they don’t want to

sell lemons — they wanted to predict which cars were

good buys and which cars were likely to be lemons.

It turned out that the auction catalogues have all these

things, like the make of the car, the model of the car, the

number of miles, how many owners. Do you want to

hazard a guess as to what the most predictive feature was?

HHmmm, I dommm, I don't hn't haavve a ce a cllue — tue — thhe coe cololor or of tf thhe ce caar?r?

Funnily enough, you're 100 percent right! It was the color

of the car. It turns out that if you group car color into

usual color and unusual color, the unusual color cars were

much more likely to be reliable than usual color cars. The

intuition there is that if you're the first buyer of an orange

car — orange was the most reliable car, incidentally —

you probably were an enthusiast because that's kind of an

odd color car and you would have had to go out of your

way to find an orange-color car. Therefore, you probably

looked after it better, so by the time it goes up for sale in a

secondhand auction, it's in much better condition than a

standard color car that somebody didn't particularly care

that much about.

In order to find that in the data you've got to use some

mathematics to discover that correlation, but really it's all

about having clever creative ideas and then testing those

ideas against the data and seeing whether they hold true

or not. That's one of the characteristics that make a great

data scientist.

WWhhaat at arre se soomme oe otthher cer chhaaracracttereriissttics yics yoouu’’vve fe foounund td thhaatt

mmaakke a ge a grreeaat dt daatta sa sciencienttiisst?t?

You have to be able to use good coding practices or

programming practices. If you set up a framework at the

beginning of a predictive modeling problem that allows

you to iterate very quickly and try lots of ideas, you'll be

able to try more ideas than somebody else that you're

competing against, and so you'll be able to discover more

in the data.

Also being tenacious. Typically in a competition, if you're

on top of the leader board and then you log in the next

day and somebody's ahead of you, you have to be

tenacious and determined enough to keep going until you

get ahead of them. And then they leapfrog you and you

have to keep working till you leapfrog them.

WWhhen a coen a commppaanny hy hososts a cots a commppetetiittioion, an, arre te thheey loy loookkiningg

ffoor jr juusst tt thhosose te toop fip fivve de daatta sa sciencienttiissts? Dts? Dooes aes annyybbooddy ey evverer

wawannt tt to uo uttiilize tlize thhosose de daatta sa sciencienttiissts tts thhaat rat rannk lok lowwer?er?

We used to think that one of the reasons that we got really

great results is because you get a lot of these offbeat ideas.

We ran a problem for NASA, and somebody in glaciology

made the first breakthrough. Now, we would not typically

turn to a glaciologist to solve a problem for NASA. What

we found, perhaps a little bit unromantically, is that the

same people keep winning or performing well again and

again and again, so there is this kind of core data science




skill that allows somebody to do well, irrespective of the

precise nature of the problem.

Kaggle's business model is basically that once we discover

who these top talents are, we put them to work on

commercial problems. We do this in one of two ways. The

first is through a private competition, so you might be a

bank, you want to build an algorithm but you can't

release PII data, data that identifies individual people. So

we can host a private competition where we'll invite five

to ten of our very best data scientists, and they compete

behind the scenes on your problem, under NDA. That

way, the sensitivity of your data is not revealed.

The other thing we do is for problems that are less well

specified, we have a partnership with EMC Greenplum

whereby any Greenplum customer can come into Kaggle

and rent one of our top data scientists for a month, for

instance, to build them a predictive modeling algorithm.

HHoow ow offtten in ten in thhe coe commppetetiittioion in is IP [ins IP [intteellleleccttuuaal pl prroopperertty]y]

ppuut int intto po prroodducucttioion?n?

It really depends. Typically, when we work with startups,

for example, or early-stage companies, the algorithms get

put into production very quickly. Often when we're

working in regulated industries, it takes a lot longer. For

example with Allstate, the first model we built for them is

a claims prediction model. Any new claims model has to

go through regulators, so there's a very long lead time

between an algorithm being developed and it being put

into production.

The way a lot of companies are set up at the moment is

you'll have one group that's responsible for data and

handling the data warehouse, and then you'll have

analysis group that's responsible for building these

algorithms, and then you'll have a third group, an IT

group, that's responsible for putting those algorithm into

production. It's sometimes challenging for the IT group

to put sophisticated algorithms into production, and so

have a service where we will host winning algorithms on

behalf of companies.

WWhhen sen shhoouuld ald an on orrgaganizanizattioion tn thinhink ak abboouut ht hososttining ag a

cocommppetetiittioion? Dn? Dooes ies it nt neeeed td to ho haavve be big dig daattaassets oets or br bigig

ccaassh? Ah? Annd wd whhaat st soorrts ots of pf prroobblemlems as arre be besest st suiuitteed fd foorr

cocommppetetiittioion?n?

I would say there are two situations where a company

should be looking at a competition as opposed to a single

data scientist. The first one is when the problem is very

valuable.

What we find with a competition is that, because we're

showing people in real time how they're performing

against others on a leaderboard, what ends up happening

is you might make an entry that you're really happy with

and, if you were working inside a company or by yourself

in isolation, you would have stopped there. But then

somebody else gets ahead of you on the leaderboard and

you keep working until you get ahead of him or her. And

they keep working until they get ahead of you.

This leapfrogging continues until the point where you

reach what we call the limit of what's possible given the

information and the dataset. Each dataset only contains

so much information and, when you put people together

in a competition and you give them feedback on a live

leaderboard, they'll drive each other to the point where

they've literally extracted everything there is to get out of

the dataset. Having a live leaderboard in front of people

means you get all the value there is to get out of a dataset.




If you're building an extremely valuable algorithm, you

want to squeeze all the juice out of it, because there could

be half a billion dollars worth of ROI in your dataset, you

want to make sure you get every little bit of it.

The second scenario to host a competition is when you

have a very challenging problem. You might be dealing

with unstructured data, such as text. Or you might be

dealing with high dimensional data or sparse data with

lots of missing values. Or social graph data. In situations

where the dataset is very difficult, we've noticed that the

difference between what a company can do themselves

and what they'll get out of our data scientists is much

bigger.

At the bottom end, competitions will give you a ten

percent lift over what you're already doing. At the top

end, it'll be in the order of three to four hundred percent.

We've noticed that generally the more complex the data,

the more challenging the data, the greater the lift.




About the Author

Renee Boucher Ferguson is aresearcher and editor at MIT SloanManagement Review.




Do You Need a Data Dictator?JEANNE ROSS (MIT SLOAN), INTERVIEWED BY DAVID KIRON AND RENEE BOUCHER FERGUSON

Many companies are suddenly dealing with petabytes of information instead of

terabytes. Keeping track of all that data and creating value from it, says Jeanne

Ross, director of the MIT Sloan Center for Information Systems Research, may

require more than technology — it may just require a data dictator.

Jeanne Ross, director of the MIT Sloan Center for Information Systems

Research

Open a newspaper or magazine, click on a news story or

case study and there it is, writ large: The promise of big

data. What has brought about this sudden interest in data

analytics?

Jeanne Ross, the director and principal research scientist

at the MIT Sloan Center for Information Systems

Research (CISR), says it is a confluence of events:

companies are getting better at the processes that create

their data, new technologies are creating new ways to

access and analyze that data and business leaders are

being bombarded with stories of others' success with

analytics.

Ross has been at MIT Sloan School of Management for

the past 19 years. For the last four years she's been the

director of CISR, which focuses on how companies get

value from information technology and on the

digitization of processes and effective use of information.

CISR has 85 corporate sponsors that fund the Center's

research, provide feedback and participate in events.

For her part, Ross examines the organizational and

performance implications related to enterprise

architecture, IT governance, business processes,

disruptive technologies and business agility in an

increasingly global, information-intensive world. She also

leads executive education courses on IT management,

working with such corporations as PepsiCo, McKinsey,



Copyright © Massachusetts Institute of Technology, 2012. All rights reserved. Reprint #54123 http://mitsmr.com/1oLuwZASPECIAL COLLECTION • “[TITLE]”• MIT SLOAN MANAGEMENT REVIEW 13

General Electric, TRW, Cisco, Commonwealth Bank of

Australia, IBM and Credit Suisse.

In a conversation with David Kiron, executive editor of

MIT Sloan Management Review Innovation Hubs, and

Renee Boucher Ferguson, a contributing editor at MITSloan Management Review, Ross discussed the issues

facing companies today as they explore the potential of

data analytics for their organizations.

HHoow do yw do yoou finu find td thhe vae vallue in yue in yoour dur daatta?a?

First of all, you have to know what is going to make you

great. If you want to run yourself as a company that is

data savvy, information savvy, analytics savvy, you need

great data about your business.

That means somebody is dictating. There is somebody

who says, "This is how we will define sales, this is how we

will define returns, this is when we will register revenue,

and we are all living by this rule. Until we do that, we

don't have data that's useful for most kinds of analytics.

We can still go out and buy demographic data and

probably learn something quite useful. But if we want to

know how to avoid stock-outs in our stores or what

products are of greatest interest to a particular customer

segment, we're going to need the data cleaned up." And

that's a major commitment. A lot of companies can't get

that done.

There are a lot of companies right now that are saying,

"Why don't we do something cool with analytics?" when

they have sloppy business processes and equally sloppy

data. And you know, they're just never going to get there.

WWhhaat it is as an exan exammpple ole of a cof a commppaanny ty thhaat ht haas ts thhe re rigighhtt

aapppprrooacach th to do daatta?a?

Aetna is a good example. In 2002, Ron Williams is

president, and he says, "Okay, we lost about $270 million

last year. Let's figure out what went wrong." He brings in

all of his senior execs, and he says, "Tell me about your

part of the business." And he said that every single line of

business showed data showing they were making a profit.

So, here's the dilemma: Everybody says their data is

showing that they made a profit, but the company is

losing a ton of money. Something's wrong.

Ron said, "The first thing I figured out is that I was going

to be the single source of truth. I was going to dictate

every piece of data, and you are going to use my

definition of data." What he got from that was totally

consistent reports, and then he knew how to guide people

through the process.

This wasn't to beat up on people. This was to get the truth

about the business so he knew what to fix. Because after

his first set of meetings, he had no idea what was broken.

It didn't look like anything was broken. This is what I

think the great leaders get right: they dictate a single

source of truth.

ThThaat's st's sucuch a gh a grreeaat exat exammpplele. B. Buut in tt in thhaat exat exammpplele, t, thherere ie iss

a cra criisisis. Is. In lieu on lieu of a crf a criisisis, ws, whhaat ct caan hn heellp tp to go get leet leaderaders ts too

rreecogcognize tnize thhaat tt thherere ne neeeedds ts to bo be a se a shifhift in tt in thhe wae way ty thheeyy

vavallue due daatta?a?

My best cases are all about a moment of truth where a

company just said well, we are going to change or we are

going to go under. A lot of them do create false crises.

We've worked a lot with Commonwealth Bank of

Australia. They didn't have a real crisis. In fact, they were,

by some measures, the largest bank in Australia. But

management defined a crisis, which was to say, "Our

operating costs are high and our customers' satisfaction




ratings are low. What this means is greatness is not

sustainable."

So they set two goals, one for operating costs and one for

customer satisfaction. Then, they looked at structurally

what was broken. And they started fixing the organization

— the processes, the technology and the data — in a

phased approach where they said, "We're just going to

keep getting better at certain things." And sure enough,

five years later, they are looking like an incredible

powerhouse.

You do have to have a burning platform. You have to have

this message that you're either under siege by competitors

or the world is changing and if you don't get on this train,

it's just going to leave you at the station.

ThThaat ct caalllls ts to mino mind td twwo po poottenenttiiaalllly coy conflicnflicttining ideg ideaas. Ons. Onee

iis ts thhe idee idea ta thhaat it it's rt's reeaalllly hy haarrd td to go get tet to bo be ne numumbber oer onnee,,

bbuut et evven hen haarrder tder to so sttaay ay at nt numumbber oer onnee. Th. The oe otthher ier is: if is: if itt

aain't bin't brrookkee, do, don't fix in't fix it. Ht. Hoow do yw do yoou cru creeaatte a be a bururninningg

ppllaattffoorrm, wm, whhen ten thherere ie is a ts a tenensiosion bn betetwweeen ten thhosose dife difffererenentt

minminddssets?ets?

That's a great contrast. Exactly right. How do you do that?

P&G was very much in this position in 2000. Results were

good but trends, particularly cost trends, were bad. They

had grown to 4 billion consumers. But the growth

markets were in developing countries. Those markets

really challenged their business model. In Turkey,

consumers often want to buy one diaper at a time. In

India, many consumers buy little, single-use shampoo

bottles. And suddenly, P&G is saying, "Wow, we want to

keep growing and our investors expect us to keep

growing, but we are really hitting some limits to our

standard processes."

That's their dilemma right now. Wall Street is saying,

"Guys, you need to grow." And yet, if you look at what

they are capable of doing, you think, "Man, a lot of

companies would kill to get there."

It's not clear how they get to the next level. What do you

do about that if you're P&G? That's really hard. Except

that you know two things. All the analysts are telling you

that Colgate and Unilever are not having the problem

you're having. That's a very interesting problem.

CCoouuld yld yoou expu expllaain tin thhe idee idea oa of sf sacracreed dd daatta ta thhaat yt yoou tu taallkkeedd

aabboouut in yt in yoour rur reecencentt e-Ce-Chhaat wt wiitthh AAlllAlAnnaallytytics.coics.comm??

First, I should tell you I learned this term from Tom

Nealon, the former CIO at Southwest Airlines [now a

board director there], and he got it from Charlie Feld,

who was the CIO at Frito-Lay and a variety of other

places. But the idea is simply that your data is not all

equal, and if you treat it as if it is all equal, you're never

going to get anywhere. So figure out first of all, what's the

single most important thing.

Tom said that at J.C. Penney's, which is where he went

after he was at Southwest, the purchases were the single

most important thing. You needed to know product,

basically. At Southwest, it had been the customer

reservation, because that runs through the entire system.

At UPS, it's the package data. By the way, UPS is brilliant

at analytics.

There are other things that matter. But you have to let the

others go until you've got the single most important data

fixed. Some of these fixes will not be as elegant as you

would like. What you'd love to do is just rip out all the

legacy that touches the customer record itself and start

from scratch, but you can only do that if you're willing to




turn it all off. Can you turn off all your systems for a year,

rebuild it? You can't.

We started learning this in the '90s. I think we were

surprised how hard it was. It was kind of like, "Yes, let me

fix my platform. Quick, put in SAP." And when that didn't

work, there was this depression on top of major financial

catastrophes. It's like, "What do we do now?" The

inclination is to say, "Well, ripping out the legacy and

replacing it with an ERP doesn't work, so we're not going

to do that." But if anybody's answering with "Oh, well,

then why don't we do data analytics instead," they are in

for a shock because they're really not going to have the

data to do the analytics — and they won't have the

discipline to implement any new processes required to

take advantage of the learning from analytics.

If your core data is bad, you can do analytics around the

edges, but you're never going to figure out how to avoid

stock-outs or better serve customers. You're not going to

figure it out because you don't have the data. There is just

some data that, as messy as the process is, you've got to

get right, and if you don't give that data all the attention it

needs, you're not going to get there.

WWhhaat do yt do yoou tu thinhink ik is rs resesppoonnsisibble fle foor tr thhe ne neew waw wavve oe off

ininttereresest in dt in daatta aa annd ad annaallytytics?ics?

Companies are starting to get better and better at storing

data and finding easy ways to get to it. So the technology

has made things possible that weren't possible before.

And companies have learned how to use that technology.

There is much more of a readiness than there was years

ago, when we learned that Capital One was doing these

incredible analytics that was making it possible for it to

customize credit cards to very small demographic

segments. Now it's like, "Wait a minute, shouldn't we all

be doing that?"

So I think there has been an announcement that analytics

has real potential at a moment when companies are

saying, "Wow, we have data that we never used to have,

and we have access to that data in ways we never dreamed

of. And then, there's all this external data that we can go

out and get. So yeah, let's analyze it. Let's get smart. Let's

do something our competitors can't do."

It feels like the moment for analytics, though it's still

much more about the promise than about the reality.

Some of the things we learn about our customers or

products or more general demographic trends have the

potential to stimulate some creative opportunities, but if a

company doesn't have a vision for how it will succeed,

these isolated findings will tend to distract rather than

drive business success. Businesses need core

competencies and they need digitized platforms if they

want to succeed in a digital economy and address the

learning from data analytics. That's the hard part of

analytics. Anyone can hire a quant jock to analyze some

data. Putting it into action requires enterprise capabilities.

WWhhen wen woouuld yld yoou su saay ty thihis ns neew waw wavve oe of inf inttereresest in dt in daatta aa anndd

aannaallytytics rics reeaalllly by begaegan?n?

The real excitement has been the last couple of years. It's

the "big data" thing. It's that suddenly we have petabytes.

What's interesting is that organizations go from terabytes

to petabytes, and both of them are unimaginable, right?

[A petabyte of data is a million times bigger than a

gigabyte and a thousand times bigger than a terabyte.]

And suddenly organizations find, "We're in petabytes.

And oh my goodness, we'd better get value from it." So

part of it is the hype. Part of it is that it's really possible to




enable our people to work smarter and to make strategic

decisions.

The hard part for a manager right now is focus. We just

don't know how much or how big or what next turn we

should take. And the sense is, "I shouldn't just sit around

and watch because other companies are getting smarter

and I'm watching." That feels like a bad formula.




About the Authors

David Kiron is executive editor ofMIT Sloan Management Review‘s

Big Ideas initiative. Renee BoucherFerguson is contributing editor ofthe Data and Analytics Innovation

Hub at MIT Sloan ManagementReview.




PDFs Reprints Permission to Copy Back Issues

Articles published in MIT Sloan Management Review are copyrighted by theMassachusetts Institute of Technology unless otherwise specified at the end of anarticle.

MIT Sloan Management Review articles, permissions, and back issues can bepurchased on our Web site: sloanreview.mit.edu or you may order through ourBusiness Service Center (9 a.m.-5 p.m. ET) at the phone numbers listed below.Paper reprints are available in quantities of 250 or more.

To reproduce or transmit one or more MIT Sloan Management Review articles byelectronic or mechanical means (including photocopying or archiving in anyinformation storage or retrieval system) requires written permission.

To request permission, use our Web site: sloanreview.mit.eduorE-mail: [email protected] (US and International):617-253-7170 Fax: 617-258-9739

Posting of full-text SMR articles on publicly accessible Internet sites isprohibited. To obtain permission to post articles on secure and/or password-protected intranet sites, e-mail your request to [email protected].


SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ® indicates USA registration. Other brand and product names are trademarks of their respective companies. 107969_S146054.1015

what makes a data scientist great? · pdf filewhat makes a data scientist great? ... a big...

Documents